www.HIMSSConference.org
#smartHIT
FEBRUARY 11, 2019
ORLANDO, FL
AI and Blockchain: Stronger Together
Sriram Vishwanath
First, a little bit about me
• Professor at UT Austin (16 years)
• PhD Stanford, Engineering and
Statistics
• MS Caltech, Engineering
• Research in AI/ML: GANs, Causal
Leaning and Generalization/transfer
learning
• Past year: building a new, scalable,
provably secure blockchain
• Founder, BrilliantMD
AI, Blockchain and the Future
Not everything is predictable!
A careful approach to AI and Blockchain
• Population àPersonalization
• I can predict it all à AI based on measured precision
• Use across all applications àTarge te d application of Blockchain
General Directions:
• Only use Blockchain when absolutely necessary
• Use AI for targeted applications, not as a general platform
• Simpler ways of accomplishing the task in a majority of cases
• Some cases clearly benefit from AI + Blockchain
Importance of a Good Methodology
Caution: Poor methodology causes “Garbage in, Garbage out”
6
Claim s
Rx
Labs
EHR
trans form
into
tensors
feature
extraction
apply algorithms
(ML and traditional)
model
Input
Actionable
Insight
Intervention
GLM
kNN
RF
Do you really need a blockchain?
*Courtesy: Do you need a blockchain?, CIBC Insights
AI/ML
• Insights – Clinical, Operational
and Financial
• Personalization of patient care
• Business process improvement
— Reduce waste, improve bottomline
• Targeted interventions
• Streamline and prioritize clinical
and non-clinical operations
• Drive rapid informed decision
making
Blockchain
• Secure multi-party transactions
• Private Auditable Ledgers, HIPAA
compliance
• Trustless consensus between
disparate players
• Real-time resolution
• Improved transparency into
datasets
• Optional: Tokenization for
incentive design
AI/ML Architecture
• Opensource tools: Scikit-learn,
Tensorflow
• Well developed and maintained
• Preprocessing essential
• Normalization and outlier detection
• Cleanup
• Active learning and feedback
• Essential to continuously improve the
model
Blockchain Architecture
• Hybrid on and off chain
architectures
• Scalability essential
• Layer 1 and Layer 2 scalability
• Security guarantees
• Many chains use proofs that offer no
security guarantees
• Privacy guarantees
• HIPAAA compliance
AI + Blockchain: Benefits
Can transform black box AI to “glass box”. Improve interpretability and transparency of results
*Courtesy: Te j Anand
AI + Blockchain: Benefits
Blockchain enables improved targeting and personalization
AI + Blockchain: Benefits
Blockchain enables compliance through Private Auditable Ledgers (PALs)
AI + Blockchain: Benefits
Blockchain enables AI-based smart contracts to be executed in real-time
*Courtesy Hacker Noon
Example AI + Blockchain Applications
• Business process efficiencies
• Claims processing
• Provider credentialing
• Prior authorization processing
• Supply chain transparency
• Improved patient engagement
• Addressing Clinical trial efficiencies
• Clinical care improvement through patient empowerment and transparency
USE-CASE: CLAIMS
DENIAL MANAGEMENT
ISSUES WITH CURRENT PROCESS
Reactive
Manual process of rework
Lack of transparency, constantly changing rules
AI + Blockchain:
AI: Streamlines/ turbo charges existing process for claims submission by RCM team
AI: Enables staff to focus on high value/important claims, Predicts missing fields
Blockchain: Identifies and highlights missing items (documents and fields) prior to
submission
Blockchain: Enables tokenization to incentivize error-free submission/prior-
authorization
OVERVIEW: BENEFITS OF COMBINING THEM
IMPROVES CLAIM SUBMISSIONS THROUGH AI
reduces errors and predicts acceptance likelihood
REDUCES REWORK & DENIAL RATES
Increases revenue and reduces costs
FRICTIONLESS PROCESS
Blockchain enables instant traceability and
containers ensure simple integration
SECURE, SOLUTIONS
Private auditable ledger ensure HIPAA compliance
and provable security guarantees
CLOSE PARTNERSHIP WITH CARECENTRIX
Summary: AI + Blockchain, Stronger Together
• Do you need a blockchain?
• Centralized, fully trusted systems do not require a blockchain
• Are you doing AI/ML Correctly?
• Do you know the measured precision of your AI/ML tool?
• For both: Is it essential for your success? A clear use-case?
• If Yes, then:
• AI + Blockchain: A match made in heaven
• Blockchain improves AI’s transparency
• Blockchain enables greater personalization in AI
• Blockchain powers AI-based smart contracts and DAO's (Decentralized Autonomous
Organizations)
• Blockchain enables compliance through private auditable ledgers
www.HealthcareMachineLearningAI.com
#smartHIT
Sriram Vishwanath
UT Austin and BrilliantMD
sriram@utexas.edu